55 lines
1.9 KiB
Bash
Executable File
55 lines
1.9 KiB
Bash
Executable File
#!/bin/bash
|
|
# llama.cpp 服务启动 + 推理任务脚本
|
|
# 用法: ./llama_inference.sh
|
|
|
|
set -e
|
|
|
|
# ==================== 配置区域 ====================
|
|
LLAMA_CMD="./llama-server -m /path/to/model.gguf --port 8080" # 启动命令
|
|
LLAMA_PORT="8080" # 服务端口
|
|
HEALTH_URL="http://localhost:${LLAMA_PORT}/health" # 健康检查地址
|
|
MAX_WAIT=300 # 最长等待秒数
|
|
|
|
# 推理任务（按顺序执行）
|
|
TASKS=(
|
|
'curl -s http://localhost:8080/completion -d "{\"prompt\": \"你好\", \"n_predict\": 50}"'
|
|
'curl -s http://localhost:8080/completion -d "{\"prompt\": \"测试\", \"n_predict\": 50}"'
|
|
)
|
|
|
|
# ==================== 启动服务 ====================
|
|
echo "[$(date '+%H:%M:%S')] 启动 llama.cpp 服务..."
|
|
$LLAMA_CMD &
|
|
LLAMA_PID=$!
|
|
echo "[$(date '+%H:%M:%S')] 服务 PID: $LLAMA_PID"
|
|
|
|
# ==================== 等待服务就绪 ====================
|
|
echo "[$(date '+%H:%M:%S')] 等待模型加载..."
|
|
WAITED=0
|
|
while ! curl -s "$HEALTH_URL" > /dev/null 2>&1; do
|
|
if [ $WAITED -ge $MAX_WAIT ]; then
|
|
echo "[$(date '+%H:%M:%S')] ❌ 等待超时，服务未就绪"
|
|
kill $LLAMA_PID 2>/dev/null || true
|
|
exit 1
|
|
fi
|
|
sleep 2
|
|
WAITED=$((WAITED + 2))
|
|
echo "[$(date '+%H:%M:%S')] 已等待 ${WAITED}s..."
|
|
done
|
|
echo "[$(date '+%H:%M:%S')] ✅ 服务已就绪"
|
|
|
|
# ==================== 执行推理任务 ====================
|
|
echo "[$(date '+%H:%M:%S')] 开始执行推理任务..."
|
|
TASK_NUM=1
|
|
for TASK in "${TASKS[@]}"; do
|
|
echo "[$(date '+%H:%M:%S')] --- 任务 $TASK_NUM ---"
|
|
eval "$TASK"
|
|
echo ""
|
|
TASK_NUM=$((TASK_NUM + 1))
|
|
done
|
|
|
|
# ==================== 清理（可选）====================
|
|
# 如果任务执行完要关闭服务，取消下面注释
|
|
# echo "[$(date '+%H:%M:%S')] 关闭服务..."
|
|
# kill $LLAMA_PID
|
|
|
|
echo "[$(date '+%H:%M:%S')] ✅ 全部完成" |